Till Toenshoff created MESOS-1220:
-------------------------------------
Summary: Make check failure on OSX - IO error: Too many open files
Key: MESOS-1220
URL: https://issues.apache.org/jira/browse/MESOS-1220
Project: Mesos
Issue Type: Bug
Affects Versions: 0.19.0
Environment: OSX 10.9.2, Apple LLVM version 5.1 (clang-503.0.40)
(based on LLVM 3.4svn)
Reporter: Till Toenshoff
Make check runs into an abort:
{noformat}
$ make check
[...]
[ OK ]
CoordinatorTest.LearnedOnOneReplica_NotLearnedOnAnother_AnotherFailsAndRecovers
(0 ms)
[----------] 21 tests from CoordinatorTest (816 ms total)
[----------] 2 tests from RecoverTest
[ RUN ] RecoverTest.RacingCatchup
F0417 21:45:21.254204 1980908304 replica.cpp:709] CHECK_SOME(state): IO error:
/private/tmp/RecoverTest_RacingCatchup_if5Cz6/.log4/LOCK: Too many open
filesFailed to recover the log
*** Check failure stack trace: ***
@ 0x10a2f9434 google::LogMessage::SendToLog()
@ 0x10a2f9963 google::LogMessage::Flush()
@ 0x10a2fcaff google::LogMessageFatal::~LogMessageFatal()
@ 0x10a2fa059 google::LogMessageFatal::~LogMessageFatal()
@ 0x109dd8479 _CheckFatal::~_CheckFatal()
@ 0x109dd8349 _CheckFatal::~_CheckFatal()
@ 0x10a1b379a mesos::internal::log::ReplicaProcess::restore()
@ 0x10a1b3241 mesos::internal::log::ReplicaProcess::ReplicaProcess()
@ 0x10a1b696b mesos::internal::log::Replica::Replica()
@ 0x1091ebd9a RecoverTest_RacingCatchup_Test::TestBody()
@ 0x10945234c
testing::internal::HandleExceptionsInMethodIfSupported<>()
@ 0x1094431ea testing::Test::Run()
@ 0x109443e72 testing::TestInfo::Run()
@ 0x1094444b0 testing::TestCase::Run()
@ 0x109449d05 testing::internal::UnitTestImpl::RunAllTests()
@ 0x109452b14
testing::internal::HandleExceptionsInMethodIfSupported<>()
@ 0x109449a39 testing::UnitTest::Run()
@ 0x10922a270 main
@ 0x7fff8a98d5fd start
@ 0x1 (unknown)
make[3]: *** [check-local] Abort trap: 6
{noformat}
That test does not fail when being run individually, hinting that we got some
general file-handle leakage problem.
The exact test that throws the abort bomb is machine and dependent. Tried it on
two MBP's and one fails a few tests earlier than the other.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
